curo1305
efc177a155
- Delete MAX_AI_CHARS constant and _client() method from anthropic_provider.py
- Add self._client = AsyncAnthropic(...) singleton in __init__ (D-07)
- Add _truncate() with 60/40 split using self._context_chars (D-13)
- Add _CLASSIFICATION_SCHEMA and _SUGGESTIONS_SCHEMA module-level constants
- classify() and suggest_topics() pass output_config with json_schema format (D-03)
- stop_reason != "end_turn" degrades to parse_classification("") (T-07-08)
- Widen __init__ signature to (api_key, model, context_chars, base_url) (uniform ctor)
- Update ai/__init__.py to pass context_chars + base_url to AnthropicProvider
- Promote test_anthropic_structured_output + add test_anthropic_stop_reason_fallback
271 lines
12 KiB
Python
271 lines
12 KiB
Python
"""
|
|
Tests for Phase 7 AI provider refactor.
|
|
|
|
Wave 2 (Plan 07-02) promotes: test_get_provider_typed, test_client_singleton,
|
|
test_generic_openai_json_mode, test_context_chars_truncation, test_smart_truncation,
|
|
test_gemini_fallback_to_parse_classification.
|
|
|
|
Wave 3 (Plan 07-03) promotes: test_anthropic_structured_output.
|
|
"""
|
|
import pytest
|
|
from unittest.mock import AsyncMock, MagicMock, patch
|
|
|
|
from ai.anthropic_provider import AnthropicProvider, _CLASSIFICATION_SCHEMA
|
|
from ai.generic_openai_provider import GenericOpenAIProvider
|
|
from ai.openai_provider import OpenAIProvider
|
|
from ai.provider_config import ProviderConfig, PROVIDER_DEFAULTS
|
|
from ai.utils import parse_classification
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Task 3: Registry-based get_provider(config: ProviderConfig) — D-06
|
|
# ---------------------------------------------------------------------------
|
|
|
|
def test_get_provider_typed():
|
|
"""get_provider() accepts a ProviderConfig and returns the correct provider class."""
|
|
from ai import get_provider
|
|
|
|
# Groq → GenericOpenAIProvider with supports_json_mode=True
|
|
config = ProviderConfig(provider_id="groq")
|
|
result = get_provider(config)
|
|
assert isinstance(result, GenericOpenAIProvider)
|
|
assert result.supports_json_mode is True
|
|
assert result._context_chars == PROVIDER_DEFAULTS["groq"]["context_chars"]
|
|
|
|
# Gemini → GenericOpenAIProvider with supports_json_mode=False (D-02)
|
|
config_gemini = ProviderConfig(provider_id="gemini")
|
|
result_gemini = get_provider(config_gemini)
|
|
assert isinstance(result_gemini, GenericOpenAIProvider)
|
|
assert result_gemini.supports_json_mode is False
|
|
|
|
# Unknown provider → ValueError
|
|
config_bogus = ProviderConfig(provider_id="bogus")
|
|
with pytest.raises(ValueError, match="Unknown AI provider"):
|
|
get_provider(config_bogus)
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Task 4: Singleton client lifecycle — D-07
|
|
# ---------------------------------------------------------------------------
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_client_singleton():
|
|
"""AsyncOpenAI is instantiated exactly once per OpenAIProvider instance (D-07)."""
|
|
with patch("ai.openai_provider.AsyncOpenAI") as mock_cls:
|
|
# mock_cls() returns a mock instance — configure chat.completions.create
|
|
mock_instance = MagicMock()
|
|
mock_instance.chat = MagicMock()
|
|
mock_instance.chat.completions = MagicMock()
|
|
mock_instance.chat.completions.create = AsyncMock(
|
|
return_value=MagicMock(
|
|
choices=[MagicMock(message=MagicMock(content='{"assigned_topics":[],"new_topic_suggestions":[]}'))]
|
|
)
|
|
)
|
|
mock_cls.return_value = mock_instance
|
|
|
|
provider = OpenAIProvider(api_key="test-key", model="gpt-4o", base_url=None, context_chars=1000)
|
|
|
|
# Call classify twice on the same instance
|
|
await provider.classify("doc text", [], "sys")
|
|
await provider.classify("doc text", [], "sys")
|
|
|
|
# AsyncOpenAI class should have been called exactly once (in __init__)
|
|
assert mock_cls.call_count == 1
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Task 4: JSON-mode conditional — D-01
|
|
# ---------------------------------------------------------------------------
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_generic_openai_json_mode():
|
|
"""GenericOpenAIProvider passes response_format only when supports_json_mode=True."""
|
|
synthetic_response = MagicMock(
|
|
choices=[MagicMock(message=MagicMock(
|
|
content='{"assigned_topics":["finance"],"new_topic_suggestions":[]}'
|
|
))]
|
|
)
|
|
|
|
# supports_json_mode=True → response_format present in call kwargs
|
|
with patch("ai.openai_provider.AsyncOpenAI") as mock_cls:
|
|
mock_instance = MagicMock()
|
|
mock_instance.chat.completions.create = AsyncMock(return_value=synthetic_response)
|
|
mock_cls.return_value = mock_instance
|
|
|
|
provider_json = GenericOpenAIProvider(
|
|
api_key="key", model="gpt-4o", base_url=None,
|
|
context_chars=1000, supports_json_mode=True
|
|
)
|
|
await provider_json.classify("text", [], "sys")
|
|
call_kwargs = mock_instance.chat.completions.create.call_args.kwargs
|
|
assert "response_format" in call_kwargs
|
|
assert call_kwargs["response_format"] == {"type": "json_object"}
|
|
|
|
# supports_json_mode=False → response_format absent from call kwargs (Gemini preset path)
|
|
with patch("ai.openai_provider.AsyncOpenAI") as mock_cls2:
|
|
mock_instance2 = MagicMock()
|
|
mock_instance2.chat.completions.create = AsyncMock(return_value=synthetic_response)
|
|
mock_cls2.return_value = mock_instance2
|
|
|
|
provider_no_json = GenericOpenAIProvider(
|
|
api_key="key", model="gemini-2.0-flash",
|
|
base_url="https://generativelanguage.googleapis.com/v1beta/openai/",
|
|
context_chars=1000, supports_json_mode=False
|
|
)
|
|
await provider_no_json.classify("text", [], "sys")
|
|
call_kwargs2 = mock_instance2.chat.completions.create.call_args.kwargs
|
|
assert "response_format" not in call_kwargs2
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Task 4: Smart truncation — D-12/D-13
|
|
# ---------------------------------------------------------------------------
|
|
|
|
def test_context_chars_truncation():
|
|
"""Provider with context_chars=100 truncates a 500-char input."""
|
|
provider = OpenAIProvider(api_key="", model="gpt-4o", base_url=None, context_chars=100)
|
|
long_text = "a" * 500
|
|
result = provider._truncate(long_text)
|
|
assert len(result) < 500
|
|
assert "[...truncated...]" in result
|
|
|
|
|
|
def test_smart_truncation():
|
|
"""_truncate uses 60% head + 40% tail of context_chars."""
|
|
provider = OpenAIProvider(api_key="", model="gpt-4o", base_url=None, context_chars=1000)
|
|
# Build a distinguishable input where head and tail chars differ
|
|
input_text = "H" * 5000 + "T" * 5000 # 10000 chars total
|
|
result = provider._truncate(input_text)
|
|
# head = int(1000 * 0.6) = 600, tail = 1000 - 600 = 400
|
|
assert result.startswith("H" * 600)
|
|
assert result.endswith("T" * 400)
|
|
assert "[...truncated...]" in result
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Task 4: Gemini fallback to parse_classification (D-02 contract enforcement)
|
|
# ---------------------------------------------------------------------------
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_gemini_fallback_to_parse_classification():
|
|
"""D-02: GenericOpenAIProvider(supports_json_mode=False) calls parse_classification()
|
|
and does NOT send response_format to the API.
|
|
"""
|
|
raw_content = '{"assigned_topics":["x"],"new_topic_suggestions":[],"reasoning":"r"}'
|
|
|
|
with patch("ai.openai_provider.AsyncOpenAI") as mock_cls:
|
|
mock_create = AsyncMock(
|
|
return_value=MagicMock(
|
|
choices=[MagicMock(message=MagicMock(content=raw_content))]
|
|
)
|
|
)
|
|
mock_instance = MagicMock()
|
|
mock_instance.chat.completions.create = mock_create
|
|
mock_cls.return_value = mock_instance
|
|
|
|
# Wrap parse_classification so we can assert it was called
|
|
with patch(
|
|
"ai.generic_openai_provider.parse_classification",
|
|
wraps=parse_classification,
|
|
) as mock_parse:
|
|
provider = GenericOpenAIProvider(
|
|
api_key="",
|
|
model="gemini-2.0-flash",
|
|
base_url="https://generativelanguage.googleapis.com/v1beta/openai/",
|
|
context_chars=8000,
|
|
supports_json_mode=False,
|
|
)
|
|
result = await provider.classify("doc text", [], "sys")
|
|
|
|
# Result must be a valid ClassificationResult
|
|
assert result.topics == ["x"]
|
|
|
|
# parse_classification was called with the raw content (D-02 contract)
|
|
assert mock_parse.called
|
|
mock_parse.assert_called_once_with(raw_content)
|
|
|
|
# response_format must NOT have been sent to the API (Gemini preset path)
|
|
call_kwargs = mock_create.call_args.kwargs
|
|
assert "response_format" not in call_kwargs
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Task 1 (Plan 07-03): AnthropicProvider output_config structured output — D-03
|
|
# ---------------------------------------------------------------------------
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_anthropic_structured_output():
|
|
"""AnthropicProvider.classify() passes output_config with the classification schema (D-03).
|
|
|
|
Verifies:
|
|
- output_config kwarg is present in the messages.create call
|
|
- output_config value matches _CLASSIFICATION_SCHEMA exactly
|
|
- Provider correctly reads response.content[0].text when stop_reason == "end_turn"
|
|
- Singleton _client is reused (AsyncAnthropic constructed once per provider instance)
|
|
"""
|
|
provider = AnthropicProvider(api_key="test-key", model="claude-sonnet-4-6", context_chars=100)
|
|
|
|
# Build a stub response that models a successful end_turn response
|
|
stub_content = MagicMock()
|
|
stub_content.text = '{"assigned_topics":[],"new_topic_suggestions":[]}'
|
|
stub_response = MagicMock()
|
|
stub_response.content = [stub_content]
|
|
stub_response.stop_reason = "end_turn"
|
|
|
|
mock_create = AsyncMock(return_value=stub_response)
|
|
|
|
with patch("ai.anthropic_provider.anthropic.AsyncAnthropic") as mock_cls:
|
|
mock_client = MagicMock()
|
|
mock_client.messages = MagicMock()
|
|
mock_client.messages.create = mock_create
|
|
mock_cls.return_value = mock_client
|
|
|
|
# Re-create provider inside the patch so self._client uses the mock
|
|
provider = AnthropicProvider(
|
|
api_key="test-key", model="claude-sonnet-4-6", context_chars=100
|
|
)
|
|
result = await provider.classify("short doc text", [], "sys prompt")
|
|
|
|
# output_config must be present and match the schema (D-03)
|
|
call_kwargs = mock_create.await_args.kwargs
|
|
assert "output_config" in call_kwargs, "output_config must be passed to messages.create()"
|
|
assert call_kwargs["output_config"] == {
|
|
"format": {"type": "json_schema", "schema": _CLASSIFICATION_SCHEMA}
|
|
}, "output_config must use _CLASSIFICATION_SCHEMA"
|
|
|
|
# AsyncAnthropic must have been constructed exactly once (D-07 singleton)
|
|
assert mock_cls.call_count == 1, "AsyncAnthropic must be constructed once (singleton)"
|
|
|
|
# Result must be a valid ClassificationResult
|
|
assert result.topics == []
|
|
assert result.suggested_new_topics == []
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_anthropic_stop_reason_fallback():
|
|
"""When stop_reason is not 'end_turn', AnthropicProvider falls back to empty ClassificationResult.
|
|
|
|
T-07-08: refusal or max_tokens stop_reason must not crash — parse_classification("") returns
|
|
an empty result rather than raising an exception.
|
|
"""
|
|
stub_content = MagicMock()
|
|
stub_content.text = '{"assigned_topics":["should","be","ignored"]}'
|
|
stub_response = MagicMock()
|
|
stub_response.content = [stub_content]
|
|
stub_response.stop_reason = "max_tokens" # simulated refusal / truncation
|
|
|
|
mock_create = AsyncMock(return_value=stub_response)
|
|
|
|
with patch("ai.anthropic_provider.anthropic.AsyncAnthropic") as mock_cls:
|
|
mock_client = MagicMock()
|
|
mock_client.messages = MagicMock()
|
|
mock_client.messages.create = mock_create
|
|
mock_cls.return_value = mock_client
|
|
|
|
provider = AnthropicProvider(api_key="k", model="claude-sonnet-4-6", context_chars=1000)
|
|
result = await provider.classify("doc text", [], "sys")
|
|
|
|
# stop_reason != "end_turn" → raw == "" → empty ClassificationResult, no crash
|
|
assert result.topics == []
|
|
assert result.suggested_new_topics == []
|